Compared to white patients with advanced cancer, black patients experience inferior pain control and less frequent and less effective discussions about symptoms, prognosis and treatment preferences. It is not clear the degree to which these differences are due to patient factors (e.g. asking fewer questions) or physician factors (e.g. implicit/unconscious racial bias, poor communication). Preliminary studies suggest that activating patients to be more assertive and to ask more questions during the visit can mitigate racial differences in communication and pain management. The proposed study addresses this critical question though a field experiment. Using novel standardized patient methodology, we will randomly assign black or white patients with identical cancer history and symptoms to primary care and oncology physicians to examine whether potentially mutable patient and physician behaviors (activation and patient-centered communication) might mitigate some of these racial differences, including those that are due to implicit biases. Aims: 1. To examine the magnitude of racial differences in physicians' assessment of pain, discussions about prognosis and treatment choices, and use of guideline-concordant pain management. 2. To examine whether patient activation mitigates racial differences in physician communication behaviors and pain management decisions. 3. To explore potential moderators of racial differences in physician communication behaviors and pain management decisions, such as patient-centered communication and implicit unconscious bias. Methods: We will conduct a randomized experiment examining mutable patient and physician behaviors. We will create 4 unannounced standardized patient (SP, secret shopper) roles portraying identical scenarios - a 70-year old patient with advanced cancer and uncontrolled pain - differing only by patient race (black/white) and level of activation (high/typical). We will randomly assign 120 physicians to see 2 unannounced covert SPs of the same race over a 12-month period, but differing according to activation. Outcome measures: We will transcribe and code audio-recorded physician-SP visits. We will code the depth of physician pain assessment using the Measure of Physician Pain Assessment, developed in our pilot (Cronbach's 1 = .68). We will assess prognosis and treatment choice communication using the Prognosis and Treatment Choices measure, which we developed in our pilot (Cronbach's 1 = 0.80). We found racial differences on both of these outcomes in our pilot. We will assess guideline-concordant pain medication recommendations using chart audit and review of transcripts. We will assess Exploring and Validating Concerns a component of patient-centered communication that has been associated with prescribing behavior, and the Implicit Associations Test, a measure of implicit bias to examine their moderating effects (Aim 3).